A Sociological Analysis of Structural Racism in Student List Products
ozanj.github.io/student_list_hsls/slides/student_list.html
Introduction
The market for college access
A two-sided matching problem in which market allocates students to colleges (Hoxby, 1997; Hoxby, 2009)
- Barriers to efficient market:
- transportation costs; information costs
- Students
- Goal: want to attend college
- Information problem: Don’t know where they will be admitted, how much it will cost
- Colleges want to enroll students
- Enrollment goals: academic profile; revenue; diversity; internal constituents
- Information problem: Don’t know who/where the “good” students are, how to contact them
Matchmaking
- “Inquiries” (student as first contact)
- Students send test scores or fill out inquiry form
- The enrollment problem
- Most colleges cannot survive/thrive solely from students who reach out on their own
- Must find desirable prospects who can be convinced to apply/enroll
Student lists products
“Student list” products are a matchmaking intermediary that connects colleges to prospects
- Third-party vendors obtain data and contact info about prospecs (e.g., Testing orgs, search engines)
- Vendors sell contact information of prospects to colleges looking for students
- Colleges choose prospect profiles by filtering on search filters (e.g., zip code, test score)
Policy concerns about student list products
- Problem with underlying products
- Search filters incorporate “racialized inputs” (Norris, 2021) that systematically disadvantage underrepresented students of color
- Problem with utilization of products
- University administrators may choose combinations of search filters that result in racial exclusion
Research questions
- What is relationship between search filters and racial composition of included vs excluded students?
- How do public universities use racialized search filters in concert with other search filters when purchasing student lists?
- What is observed racial composition of student list purchases that utilize racialized search filters in concert with other search filters?
Research overview
Research questions
- What is relationship between search filters and racial composition of included vs excluded students?
- Reconstruct College Board student list product using nationally representative sample of 9th graders from 2009 (NCES High School Longitudinal Survey)
- Simulate which students are included versus excluded when certain search filters are utilized
- How do public universities use racialized search filters in concert with other search filters when purchasing student lists?
- Analyze 830 student lists purchased by 14 public universities, collected via public records requests
- What is observed racial composition of student list purchases that utilize racialized search filters in concert with other search filters?
- Analyze targeted student list purchases, where we obtained both the order details and deidentified prospect-level data
Literature Review
Scholarship on recruiting from sociology
Enrollment funnel: prospects >> leads >> inquires >> applicants >> admits >> enrolled
- Most scholarship on enrollment management focuses on latter stages (admissions, financial aid)
- Body of research in sociology that analyzes recruiting “in the wild”
Recruiting from perspective of high school students (Holland, 2019)
- Underrepresented students sensitive to feeling “wanted” by colleges
Connections between high schools and colleges from organizational behavior perspective
- Off-campus recruiting visits indicate a network tie and enrollment priorities
- recruiting from perspective of private college (Stevens, 2007), private HS counselors (Khan, 2011)
- Recruiting visits by public research univs (e.g., Salazar, Jaquette, and Han, 2021; and Salazar, 2022)
Recruiting at open-access PSIs for adults (e.g., and Cottom, 2017; and Posecznick, 2017)
- For-profits have demand in Black/Latinx communities because traditional colleges ignore them
Scholarship assumes that recruiting is something done wholly done by individual colleges
- Third-party products and vendors structure recruiting behavior by colleges
- How do products incorporate structural inequality? How are they utilized by colleges?
Background
Where do student lists fit in recruiting
Prospects
- Population of desirable potential students
Leads
- Prospects whose contact info has been obtained
Inquiries
- Prospects who have contacted the institution
- Institution as first contact (leads)
- Student as first contact
Interventions along the funnel
- Convert prospects to leads
- purchase student lists
- Convert leads/inquiries to applicants
- Email, mail, targeted social media
- Convert admits to enrolles
- Financial aid packages
College Board and ACT student list products
Sources of student list data
- Test takers (e.g., PSAT, SAT, AP), pre-test questionnaire (demographics, preferences)
- More recently, from college search engines (e.g., College Board Big Future)
- Students can opt in or out
What information does a list contain ([College Board template](https://drive.google.com/file/d/1Qvc_QRi9izEF1W78Lh4nNi5NsXjCZqUE))
- Contact, demographic, college preferences, limited academic achievement
Pricing
- Historically, price-per-prospect (e.g., $0.50 per name); now a subscription model
Buying lists: “search filters” control which prospects included in purchase
- Commonly used search filters (Link to ACT filters)
- e.g., HS GPA, test score range, gender, race, geography (e.g., state), enrollment probability
- Salazar, Jaquette, and Han (2022) categorizes College Board search filters into four buckets:
- Geographic
- Academic
- Demographic
- Student preferences
Filters used in 830 lists purchased by 14 public universities
College Board Search and student outcomes
Howell, Hurwitz, Mabel et al. (2021)
also see: Smith, Howell, and Hurwitz (2022) for effect of university purchasing profile on college choice
Conceptual Framework
Sociology of race
Selection devices allocate individuals to categories based on input factors (Hirschman and Bosk, 2020)
- Discretionary selection devices
- Decisions incorporate judgment of individual evaluators (e.g., holistic admissions)
- Standardized selection devices
- decisions based on mathematical function in which input values determine outcome value
- “actuarial” selection devices predict outcomes based on analysis of past cases
- Student list products are discretionary selection devices
- administrator chooses prospects by selecting search filters (inputs)
- inputs factors include descriptive characteristics and predictive analytics
Selection devices and racial inequality
- Standardized selection devices
- May reduce racial inequality if source of inequality is bias from individual decision-maker
- Do not reduce inequality due to structural racism (Bonilla-Silva, 1997)
- “colorblind” selection devices reproduce inequality by utilizing inputs related to race
- Discretionary selection devices
- Sensitive to bias from decision-makers and inputs related to race
Racialized inputs
Structural racism (Tiako, South, and Ray, 2021, p. 1143)
- “systematic racial bias embedded in the ‘normal’ functions of laws and social relations,” whereby processes viewed as neutral or common-sense systematically disadvantage marginalized groups
Racialized inputs (Norris, 2021)
- Inputs that are “theoretically and empirically correlated with historical racial disadvantage” (p. 5)
- reconstructs inputs used by Moody’s city government credit rating algorithm
Geography inputs
- US racially segregated due to historic and current laws/policies/practices
- Zip code an input in algorithm that predicts crime recidivism (O’Neil, 2016; Benjamin, 2019)
- Zip code a search filter on CB/ACT student list products
- Geodemographic classifications classify localities by consumer behavior (Leyshon and Thrift, 1999)
- CB Geodemographic filters classify census tracts by past college enrollment (College Board, 2011)
Inputs based on predictive analytics
- Analyze determinants of an outcome in past cases, use results as an input to select future cases (e.g., ACT probability of enrollment filter)
RQ1: relationship between filters and racial composition
Standardized college entrance exams and AP exams as racialized inputs
- Differences in test-taking and scores by race a function of historic/contemporary school segregation, differences in school funding and access to curricula (e.g., Reardon, Kalogrides, and Shores, 2019; Rodriguez and Hernandez-Hamed, 2020)
- Concerns about bias in question items (e.g., and Freedle, 2003; Santelices and Wilson, 2010)
P1: The condition of taking standardized assessments is associated with racial disparities in who is included versus excluded in student list products.
P2: As test score threshold increases, the proportion of underrepresented minority students included in student lists declines relative to the proportion who are excluded.
P3. As purchases filter on more affluent geographic localities (e.g., zip codes), the proportion of underrepresented minority students included in student lists declines relative to the proportion who are excluded.
Filtering on multiple racialized inputs has compounding effect on racial inequality
Methods
Data
HSLS09 & Student List Project
High School Longitudinal Study of 2009 (HSLS09)
- Unweighted analysis sample (n=16,530)
- Includes students who completed 2012/2013 survey follow-ups and obtained high school transcript data
- Weighted analysis sample represent approximately 4.2 million U.S. 9th graders in 2009
Student List Project
- Issued public records requests for student list data (2016-2020) to all public universities in four states (CA, IL, MN, TX)
- Target student list vendors: College Board, ACT
- For each purchased list, sought two pieces of data
Summary of data received
| State | # received order summary | # no order summary | # received list | # no list | # received both | # did not receive both |
|---|---|---|---|---|---|---|
| CA | 9 | 23 | 13 | 19 | 9 | 23 |
| IL | 9 | 3 | 9 | 3 | 8 | 4 |
| TX | 15 | 20 | 16 | 19 | 10 | 25 |
Orders and prospects purchased
Research design
Variables & Analyses
RQ1: What is the relationship between student list search filters and the racial composition of students who are included versus excluded from College Board student list purchases?
RQ2:
- Dependent Variable: student race/ethnicity
- Independent Variables: measures of student list academic and geographic filters
- Dichotomous measures of SAT, PSAT, AP completion
- Threshold measures of highest SAT, PSAT, AP scores, and GPA
- Economic measures of students’ zipcode, county, CBSA from ACS
- Analyses: descriptive statistics
- All analyses compare the racial composition of included versus excluded prospects when particular filters and/or filter thresholds are utilized to purchase prospect profiles
Results
RQ1
Test Takers
P1: Racial disparities in test-taking
Test thresholds
P2: SAT, PSAT score thresholds and racial composition
Test thresholds
P2: AP sScore thresholds and racial composition
Geography
P3: Zip code affluence and racial composition
Geography
Zip vs county, affluence percentiles within CBSA
Academic & Geographic
GPA (3.0+) and SAT or PSAT (across score thresholds)
Academic & Geographic
GPA (3.0+), PSAT (150+) or SAT (1050+), and Zip (across income thresholds)
Academic & Geographic
GPA (3.0+) and AP (across score thresholds)
RQ2
Broad patterns
Filters used in order purchases
Combination of filters
Filter combos used in order purchases
| Research | MA/doctoral | ||||
|---|---|---|---|---|---|
| Filters | Count | Percent | Filters | Count | Percent |
| HS grad class, GPA, SAT, PSAT, Rank, State, Race | 39 | 10% | HS grad class, GPA, SAT, Zip code | 206 | 45% |
| HS grad class, PSAT, State | 27 | 7% | HS grad class, GPA, PSAT, Zip code | 145 | 32% |
| HS grad class, GPA, PSAT, State, Race | 20 | 5% | HS grad class, SAT, State | 31 | 7% |
| HS grad class, PSAT, State, Low SES | 20 | 5% | HS grad class, GPA, SAT, PSAT, Zip code | 28 | 6% |
| HS grad class, GPA, PSAT, State | 17 | 5% | HS grad class, GPA, SAT, State | 7 | 2% |
| HS grad class, GPA, SAT, State | 16 | 4% | HS grad class, SAT, Geomarket | 6 | 1% |
| HS grad class, GPA, AP score, Geomarket | 15 | 4% | HS grad class, GPA, SAT, County | 5 | 1% |
| HS grad class, GPA, SAT, PSAT, State, Segment, Gender | 13 | 3% | HS grad class, GPA, SAT, PSAT, County | 4 | 1% |
| HS grad class, PSAT, Geomarket | 12 | 3% | HS grad class, GPA, PSAT, State | 2 | 0% |
| HS grad class, SAT, State, Low SES, College size | 11 | 3% | HS grad class, SAT, Geomarket, College type | 2 | 0% |
RQ3
Characteristics by filters
Prospect characteristics across individual filter criteria
| Academic | Geographic | Demographic | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All domestic | GPA | PSAT | SAT | HS rank | AP score | Zip code | State | Geomarket | Segment | CBSA | Race | Gender | ||||
| Total | 3,547,620 | 1,101,266 | 1,812,447 | 971,237 | 146,660 | 75,479 | 165,924 | 1,173,678 | 1,056,951 | 186,519 | 146,313 | 279,626 | 39,546 | |||
| Location | ||||||||||||||||
| % In-state | 38 | 62 | 30 | 54 | 83 | 42 | 98 | 48 | 17 | 15 | 4 | 59 | 6 | |||
| % Out-of-state | 62 | 38 | 70 | 46 | 17 | 58 | 2 | 52 | 83 | 85 | 96 | 41 | 94 | |||
| Race/ethnicity | ||||||||||||||||
| % White | 48 | 45 | 50 | 47 | 51 | 17 | 43 | 42 | 57 | 51 | 53 | 25 | 47 | |||
| % Asian | 16 | 15 | 17 | 15 | 10 | 7 | 13 | 18 | 13 | 27 | 28 | 5 | 38 | |||
| % Black | 5 | 7 | 4 | 7 | 8 | 17 | 8 | 5 | 4 | 3 | 2 | 11 | 1 | |||
| % Latinx | 21 | 24 | 19 | 22 | 23 | 46 | 27 | 24 | 16 | 11 | 8 | 46 | 6 | |||
| % AI/AN | 1 | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 0 | 0 | 0 | 2 | 0 | |||
| % NH/PI | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |||
| % Multiracial | 5 | 5 | 5 | 5 | 5 | 10 | 4 | 6 | 5 | 5 | 5 | 9 | 5 | |||
| % Other | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |||
| % No response | 4 | 3 | 3 | 3 | 2 | 1 | 4 | 3 | 4 | 3 | 3 | 2 | 3 | |||
| % Missing | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | 0 | |||
| Gender | ||||||||||||||||
| % Male | 34 | 19 | 37 | 18 | 0 | 3 | 46 | 24 | 48 | 6 | 0 | 11 | 0 | |||
| % Female | 36 | 23 | 40 | 20 | 1 | 15 | 54 | 27 | 52 | 9 | 0 | 12 | 33 | |||
| % Other | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |||
| % Missing | 30 | 58 | 22 | 63 | 99 | 82 | 0 | 49 | 0 | 85 | 1 | 77 | 67 | |||
| Household income | ||||||||||||||||
| Median income | $107K | $105K | $108K | $105K | $99K | $90K | $97K | $105K | $107K | $130K | $135K | $94K | $127K | |||
| Locale | ||||||||||||||||
| % City | 27 | 27 | 27 | 26 | 26 | 31 | 31 | 30 | 23 | 24 | 22 | 29 | 26 | |||
| % Suburban | 44 | 47 | 44 | 48 | 53 | 40 | 42 | 42 | 46 | 54 | 57 | 47 | 49 | |||
| % Rural - Fringe | 22 | 20 | 22 | 20 | 15 | 23 | 19 | 22 | 23 | 19 | 19 | 19 | 23 | |||
| % Rural - Distant | 6 | 6 | 5 | 6 | 6 | 5 | 7 | 6 | 6 | 2 | 1 | 6 | 2 | |||
| % Rural - Remote | 1 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | 0 | |||
| % Missing | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |||
Geodemographic segment filters
Filter by neighborhood segments
| 2011 D+ Cluster | SAT Math | SAT CR | Going Out of State | Percent NonWhite | Need Financial Aid | Med Income |
|---|---|---|---|---|---|---|
| 51 | 546.00 | 533.00 | 32% | 30% | 57% | $95,432 |
| 52 | 480.00 | 470.00 | 30% | 58% | 71% | $63,578 |
| 53 | 561.00 | 544.00 | 32% | 50% | 55% | $92,581 |
| 54 | 458.00 | 443.00 | 25% | 83% | 76% | $38,977 |
| 55 | 566.00 | 565.00 | 52% | 24% | 63% | $71,576 |
| 56 | 420.00 | 411.00 | 29% | 93% | 66% | $35,308 |
| 57 | 541.00 | 519.00 | 52% | 47% | 43% | $67,394 |
| 58 | 533.00 | 489.00 | 28% | 87% | 69% | $68,213 |
| 59 | 561.00 | 562.00 | 52% | 24% | 74% | $54,750 |
| 60 | 589.00 | 590.00 | 63% | 37% | 36% | $104,174 |
| 61 | 585.00 | 567.00 | 51% | 30% | 40% | $123,858 |
| 62 | 596.00 | 595.00 | 67% | 24% | 72% | $59,824 |
| 63 | 548.00 | 541.00 | 39% | 23% | 65% | $69,347 |
| 64 | 466.00 | 466.00 | 48% | 34% | 29% | $49,829 |
| 65 | 440.00 | 433.00 | 23% | 93% | 78% | $45,081 |
| 66 | 499.00 | 492.00 | 20% | 12% | 76% | $50,453 |
| 67 | 519.00 | 501.00 | 27% | 53% | 59% | $60,960 |
| 68 | 552.00 | 558.00 | 52% | 35% | 65% | $57,902 |
| 69 | 534.00 | 521.00 | 37% | 19% | 65% | $88,100 |
| 70 | 613.00 | 598.00 | 65% | 29% | 61% | $86,381 |
| 71 | 405.00 | 408.00 | 39% | 97% | 68% | $42,661 |
| 72 | 399.00 | 397.00 | 31% | 87% | 47% | $32,708 |
| 73 | 528.00 | 514.00 | 29% | 42% | 62% | $90,849 |
| 74 | 433.00 | 435.00 | 29% | 84% | 79% | $44,065 |
| 75 | 459.00 | 457.00 | 28% | 85% | 72% | $50,421 |
| 76 | 514.00 | 509.00 | 27% | 38% | 64% | $61,332 |
| 77 | 502.00 | 492.00 | 26% | 18% | 75% | $62,372 |
| 78 | 594.00 | 578.00 | 56% | 26% | 39% | $134,400 |
| 79 | 550.00 | 551.00 | 57% | 32% | 74% | $40,909 |
| 80 | 534.00 | 527.00 | 39% | 39% | 65% | $49,877 |
| 81 | 491.00 | 483.00 | 27% | 57% | 72% | $63,030 |
| 82 | 496.00 | 491.00 | 29% | 21% | 75% | $53,465 |
| 83 | 500.00 | 490.00 | 19% | 26% | 71% | $49,335 |
| Total | 512.00 | 502.00 | 32% | 43% | 65% | $70,231 |
Filter by high school segments
| 2011 D+ Cluster | SAT Math | SAT CR | Going Out of State | Percent NonWhite | Need Financial Aid | Med Income |
|---|---|---|---|---|---|---|
| 51 | 462.00 | 457.00 | 14% | 33% | 68% | $40,918 |
| 52 | 489.00 | 496.00 | 81% | 99% | 77% | $64,730 |
| 53 | 471.00 | 484.00 | 28% | 38% | 62% | $60,833 |
| 54 | 376.00 | 371.00 | 33% | 96% | 38% | $38,146 |
| 55 | 489.00 | 481.00 | 39% | 46% | 44% | $71,845 |
| 56 | 536.00 | 508.00 | 73% | 43% | 49% | $63,967 |
| 57 | 434.00 | 435.00 | 29% | 82% | 79% | $48,301 |
| 58 | 592.00 | 577.00 | 51% | 27% | 32% | $104,509 |
| 59 | 499.00 | 489.00 | 19% | 18% | 74% | $47,685 |
| 60 | 523.00 | 549.00 | 23% | 30% | 33% | $70,175 |
| 61 | 485.00 | 370.00 | 33% | 89% | 9% | $61,385 |
| 62 | 474.00 | 473.00 | 34% | 92% | 67% | $55,515 |
| 63 | 440.00 | 427.00 | 28% | 86% | 72% | $49,238 |
| 64 | 606.00 | 542.00 | 37% | 89% | 57% | $81,911 |
| 65 | 515.00 | 503.00 | 28% | 43% | 65% | $72,692 |
| 66 | 498.00 | 515.00 | 37% | 37% | 73% | $60,272 |
| 67 | 526.00 | 546.00 | 48% | 41% | 69% | $71,279 |
| 68 | 541.00 | 540.00 | 41% | 26% | 62% | $79,260 |
| 69 | 390.00 | 395.00 | 36% | 92% | 74% | $43,391 |
| 70 | 595.00 | 581.00 | 56% | 33% | 48% | $105,721 |
| 71 | 400.00 | 412.00 | 57% | 98% | 80% | $43,137 |
| 72 | 528.00 | 544.00 | 35% | 25% | 64% | $70,018 |
| 73 | 451.00 | 438.00 | 24% | 89% | 76% | $48,406 |
| 74 | 654.00 | 579.00 | 76% | 80% | 46% | $59,089 |
| 75 | 514.00 | 502.00 | 31% | 20% | 71% | $72,850 |
| 76 | 600.00 | 584.00 | 72% | 50% | 28% | $90,265 |
| 77 | 595.00 | 508.00 | 64% | 75% | 39% | $39,490 |
| 78 | 473.00 | 468.00 | 48% | 43% | 22% | $56,703 |
| 79 | 594.00 | 585.00 | 61% | 26% | 71% | $65,180 |
| Total | 514.00 | 502.00 | 32% | 44% | 65% | $70,223 |
Segment filter prospects by metro
Segment filter prospects interactive map
Women in STEM
Women in STEM prospects by metro
Targeting URM students
Race and ethnicity variables, aggregated vs. alone
Purchased profiles for students of color by metro
Purchased profiles for students of color interactive map
Zip code & test score filters
Los Angeles prospects from top income decile zip codes
Discussion
Data as capital, obfuscation, and policy research
Student list data derived from user-data of students laboring on platforms
- Marx (1978): formula for economic capital is \(M - C - M'\)
- money ($M$); commodities ($C$)
- Data as capital (Sadowski, 2019)
- Data an input into production commodities (e.g., software predicting hospital staff needs)
- Data are a commodity extracted from labor of people using digital platforms
- College Board follows \(M - C - M'-C-M''\): Invest money ($M$) to develop tests ($C$); sold to households ($M’$) yielding student list data ($C$); sold to universities ($M’’$)
- Emerging trend: wrap student list data within a software-as-service recruiting product
Obfuscation (Cottom, 2020; Pasquale, 2015)
- Opacity of digital platforms is deliberate strategy to manage regulatory environments
- Really hard to collect data about student list products or “student success” products
Policy
- Policy should regulate products sold to schools, universities, and students
- Developing regulations requires on a body of research
- Education researchers must interrogate third-party products and vendors
- Focus on structural inequality embedded in product design
References
[1] R. Benjamin. Race after technology: Abolitionist tools for the new Jim code. Medford, MA: Polity, 2019. ISBN: 9781509526390 (hardback) 9781509526406 (paperback).
[2] E. Bonilla-Silva. “Rethinking racism: Toward a structural interpretation”. In: American Sociological Review 62.3 (1997), pp. 465-480. ISSN: 0003-1224. DOI: 10.2307/2657316. URL: <Go to ISI>://WOS:A1997XG45300007.
[3] College Board. Segment Analysis Service: An educationally relevant geodemographic tagging service. Tech. rep. College Board, 2011. URL: https://secure-media.collegeboard.org/mSSS/media/pdf/segment-analysis-service-overview.pdf.
[4] T. M. Cottom. Lower ed: The troubling rise of for-profit colleges in the new economy. New Press, The, 2017.
[5] T. M. Cottom. “Where platform capitalism and racial capitalism meet: The sociology of race and racism in the digital society”. In: Sociology of Race and Ethnicity 6.4 (2020), pp. 441-449. ISSN: 2332-6492. DOI: 10.1177/2332649220949473. URL: https://doi.org/10.1177/2332649220949473.
[6] D. Hirschman and E. A. Bosk. “Standardizing biases: Selection devices and the quantification of race”. In: Sociology of Race and Ethnicity 6.3 (2020), pp. 348-364. DOI: 10.1177/2332649219844797. URL: https://journals.sagepub.com/doi/abs/10.1177/2332649219844797.
[7] M. M. Holland. Divergent paths to college: Race, class, and inequality in high schools. Rutgers University Press, 2019. DOI: 10.36019/9780813590288.
[8] J. Howell, M. H. Hurwitz, Z. Mabel, et al. Participation in student search service is associated with higher college enrollment and completion. Tech. rep. College Board, 2021. URL: https://cbsearch.collegeboard.org/pdf/college-outreach-and-student-outcomes.pdf.
[9] C. M. Hoxby. How the changing market structure of U.S. higher education explains college tuition. Tech. rep. Working Paper 6323. National Bureau of Economic Research, 1997. URL: http://www.nber.org/papers/W6323.pdf.
[10] C. M. Hoxby. “The changing selectivity of American colleges”. In: Journal of Economic Perspectives 23.4 (2009), pp. 95-118. ISSN: 0895-3309. URL: <Go to ISI>://000272677100005.
[11] S. R. Khan. Privilege: The making of an adolescent elite at St. Paul's School. Princeton studies in cultural sociology. Princeton, N.J.: Princeton University Press, 2011, p. 232. ISBN: 9780691145280 (alk. paper) 0691145288 (alk. paper) 9780691156231 (pbk.) 0691156239 (pbk.).
[12] A. Leyshon and N. Thrift. “Lists come alive: electronic systems of knowledge and the rise of credit-scoring in retail banking”. In: Economy and Society 28.3 (1999), pp. 434-466. ISSN: 0308-5147. DOI: 10.1080/03085149900000013. URL: https://doi.org/10.1080/03085149900000013.
[13] K. Marx. “Capital”. In: The Marx-Engels Reader. Ed. by R. C. Tucker. second. New York: W.W. Norton and Company, 1978.
[14] D. Norris. “Embedding racism: City government credit ratings and the institutionalization of race in markets”. In: Social Problems (2021). ISSN: 0037-7791. DOI: 10.1093/socpro/spab066. URL: https://doi.org/10.1093/socpro/spab066.
[15] C. O'Neil. Weapons of math destruction: how big data increases inequality and threatens democracy. First edition.. New York: Crown, 2016, p. x, 259 pages. ISBN: 9780553418811 (hardcover) 9780553418835 (softcover).
[16] F. Pasquale. The black box society: the secret algorithms that control money and information. Cambridge: Harvard University Press, 2015, p. 311. ISBN: 9780674368279.
[17] A. Posecznick. Selling hope and college merit, markets, and recruitment in an unranked school. Ithaca ; London: Cornell University Press, 2017. ISBN: 9781501707582 9781501709821. URL: https://muse.jhu.edu/book/52616.
[18] J. Sadowski. “When data is capital: Datafication, accumulation, and extraction”. In: Big Data & Society 6.1 (2019), p. 12. ISSN: 2053-9517. DOI: 10.1177/2053951718820549. URL: <Go to ISI>://WOS:000455488000001.
[19] K. G. Salazar. “Recruitment Redlining by Public Research Universities in the Los Angeles and Dallas Metropolitan Areas”. In: The Journal of Higher Education (2022), pp. 1–37. DOI: 10.1080/00221546.2021.2004811. URL: https://doi.org/10.1080/00221546.2021.2004811.
[20] K. G. Salazar, O. Jaquette, and C. Han. “Coming Soon to a Neighborhood Near You? Off-Campus Recruiting by Public Research Universities”. In: American Educational Research Journal 58.6 (2021), pp. 1270-1314. DOI: 10.3102/00028312211001810. URL: https://journals.sagepub.com/doi/abs/10.3102/00028312211001810.
[21] K. G. Salazar, O. Jaquette, and C. Han. Geodemographics of student list purchases: a first look. Tech. rep. Washington, DC: TICAS, 2022. URL: https://ticas.org/wp-content/uploads/2022/09/Geodemographics-of-Student-List-Purchases_A-First-Look.pdf.
[22] M. L. Stevens. Creating a class: College admissions and the education of elites. Cambridge, MA: Harvard University Press, 2007, p. 308. ISBN: 9780674026735 (alk. paper) 067402673X (alk. paper).
[23] M. J. N. Tiako, E. South, and V. Ray. “Medical Schools as Racialized Organizations: A Primer”. In: Annals of Internal Medicine 174.8 (2021), pp. 1143-1144. DOI:
Appendix
RQ1
Test Thresholds
GPA
RQ2
Academic filters
GPA filter used
SAT filter used
PSAT filter used
Geographic filters
State filter used by research universities, out-of-state
State filter used by research universities, in-state
Demographic filters
Race filter
